Positional Variance Profiles (PVPs): A New Take on the Speed-Accuracy Trade-off

PVPs display two phases; we described the second phase with a theoretical model, and the first phase empirically only [16].

The method of PVPs can be applied to several protocols including ones where there is no pre-specified width such as dual minimization protocols and is also more expressive than Fitts' law, because it characterizes whole trajectories, and not just endpoints (see Fig. 1). Another nice property of PVPs is that Fitts' law can essentially be derived from them.

Hence, by transposing the method of PVPs in HCI, we also alleviate some drawbacks associated with Fitts' law:

• Fitts' law is expressed as the interaction between the two phases of the PVP. As a result, Fitts' law parameters can be computed from the PVP parameters and can be interpreted within the information-transmission scheme.

This provides new arguments to the aforementioned unresolved problems with Fitts' law experiments.

• We propose a new protocol, amenable to the method of PVPs, which does not try to enforce endpoint width constraints to generate data, thereby removing the need for a post-hoc correction. The protocol is also simpler to conduct, since it does not require crossing D and W: a single experimental condition is enough.

Currently, however, the method of PVPs can not be applied off-the-shelf to most HCI evaluation opportunities: the method was described for 1D data only, whereas most situations in HCI feature two-dimensional pointing. A similar problem affected Fitts' law experiments, which was also initially described in 1D only, and which attracted significant effort [1,21,28,31,39]. Thus, in this work we extend the one dimensional method of PVP to so-called 2D-PVPs. In doing so, we make the connection to existing literature on 2D Fitts' law whenever that is possible. Further, the effect of